Monolayer iron oxide film on platinum promotes low temperature CO oxidation.
Y.-N. Sun, Z.-H. Qin, M. Lewandowski, E. Carrasco, M. Sterrer,
S. Shaikhutdinov*, H.-J. Freund
Abteilung Chemische Physik, Fritz-Haber-Institut der Max-Plank-Gesellschaft,
Faradayweg 4-6, Berlin 14195
Abstract. CO oxidation on a clean Pt(111) single crystal and thin iron oxide films grown on
Pt(111) was studied at different CO:O2 ratios (between 1:5 and 5:1) and partial pressures up
to 60 mbar at 400 – 450 K. Structural characterization of the model catalysts was performed
by scanning tunneling microscopy, low energy electron diffraction, Auger electron
spectroscopy and temperature programmed desorption. It is found that monolayer FeO(111)
films grown on Pt(111) are much more active than clean Pt(111) and nm-thick Fe 3 O4(111)
films at all reaction conditions studied. Post-characterization of the catalysts revealed that at
CO:O2 >1 the FeO(111) film dewets the Pt surface with time, ultimately resulting in highly
dispersed iron oxide particles on Pt(111). The film dewetting was monitored in situ by
polarisation-modulated infrared reflection absorption spectroscopy. The reaction rate at 450
K exhibited first order for O2 and non-monotonously depended on CO pressure. In O2-rich
ambient the films were enriched with oxygen while maintaining the long range ordering.
Based on the structure-reactivity relationships observed for the FeO/Pt films, we propose
that the reaction proceeds through the formation of a well-ordered, oxygen-rich FeOx (1 < x
< 2) film that reacts with CO through the redox mechanism. The reaction induced dewetting
in fact deactivates the catalyst. The results may aid in our deeper understanding of reactivity
of metal particles encapsulated by thin oxide films as a result of strong metal support
interaction.
Keywords: platinum, iron oxides, CO oxidation, strong metal support interaction.
Corresponding author: [email protected]
1
1. Introduction
The nature of an oxide support plays an important role in catalytic reactions over
highly dispersed noble metal catalysts. At a given particle size, the support effects are often
rationalized in terms of reactions taking place at the metal/oxide interface or involving
spillover of reactive species onto/from an oxide support. These mechanisms are usually
considered for transition metal oxides such as titania, ceria, etc. On the other hand, metals
supported on the reducible oxides often exhibit so-called strong metal-support interaction
(SMSI) [1-6], which in many cases manifests itself by encapsulation (decoration) of the
metal particles by a thin layer stemming from the oxide support. Formation of thin titania
layers on Pt and Pd particles upon heating in vacuo has recently been investigated by high
resolution scanning tunnelling microscopy (STM) [7-9]. It is rather obvious that the
decoration will suppress catalytic reactions occurring on metal surfaces and be particularly
detrimental for structure sensitive reactions. On the other hand, the reversal of the SMSI
state has been observed after CO hydrogenation reactions on Pt/TiO2 [10,11]. Also, partial
recovering of CO uptake (as a test for the metal encapsulation) upon oxidative treatments of
Pt/TiO2 (110) and Pt/CeO2(111) model catalysts has been reported [12,13]. Reversible
wetting-dewetting behaviour induced by ambient conditions has recently been demonstrated
by in situ STM on vanadia overlayer on Pd(111) [14]. Therefore, reaction conditions may
affect the surface structure and hence the reactivity of catalysts that underwent the
encapsulation during catalyst’s preparation.
We have recently shown that Pt supported on iron oxide Fe 3O4 (111) films also
exhibits an SMSI effect via encapsulation which is driven by the high adhesion energies
between Pt and iron oxide [15,16]. In particular, top facets of the Pt particles heated above
800 K in vacuum revealed the structure that is very similar to an ultrathin FeO(111) film
grown on a Pt(111) single crystal [17]. This finding suggests that the FeO(111)/Pt(111) film
is a suitable model system for studying the behaviour of the encapsulated Pt particles.
The adsorption properties of FeO(111) as well as of Fe 3 O4 (111) and Fe2 O3 (0001)
films under ultrahigh vacuum conditions were reviewed by Weiss and Ranke [18]. Under the
conditions typically used in temperature programmed desorption/reaction (TPD, TPR)
studies, the FeO films were essentially inert towards molecular adsorption (carbon
2
monoxide, water, ethylbenzene and styrene). Therefore, the adsorption studies were
primarily focused on surfaces of magnetite (Fe 3O4 ) and hematite (Fe 2 O3) (see also [19,20]).
The inertness of the FeO(111) film is believed to be related to the fact that the surface
exposes a close packed layer of oxygen (the monolayer film stacks as O-Fe-Pt-Pt…).
However, it is not obvious per se that the film will maintain the structure upon exposure to
realistic reaction conditions. It has been demonstrated that adsorbates, molecules, metal
atoms and clusters on thin oxide films supported on metallic substrates may, under certain
conditions, induce electron transfer through the film onto the adsorbate [21-27]. Therefore, it
is conceivable, that under favorable conditions such electron transfer in turn induces
reactivity between molecules. Formation of an appropriate transition state with high
probability is almost certainly enhanced under high-pressure conditions. If the chemical
potential, set by the gas pressure, is such that one of the reacting components is oxygen,
prone to exchange with the oxygen in the film, then restructuring of the film, accompanied
by unexpected reactivity, could be observed.
In attempts to shed light on the possible effects of elevated pressures and
temperatures on structure and reactivity of thin oxide films, we have recently initiated “highpressure” studies of FeO/Pt films with respect to CO oxidation. To combine both, reactivity
studies at atmospheric pressure and a structural control of model catalysts, we used a socalled “high-pressure” cell located inside an ultra-high vacuum (UHV) chamber equipped
with different surface-science tools. The sample, having been prepared and characterized
under UHV conditions, is sealed in the cell used as a reactor that allows using gas
chromatography (GC) analysis of products. In particular, we have found that the FeO films
in 40 mbar of CO + 20 mbar O2 at 450 K show much higher CO 2 production than the clean
Pt(111) surface [28]. Tentatively, this unexpected result has been explained on the basis of
reaction induced dewetting of the oxide film, resulting in highly dispersed FeOx
nanoparticles on Pt(111), thus effectively forming “inverted” catalysts [29].
In this paper, we report an extended study of the structure and reactivity of FeO
films with respect to CO oxidation at different CO:O 2 ratios, partial pressures and reaction
temperatures. The results suggest that the unusual activity of the ultra -thin FeO films is
intimately connected to the formation of an oxygen-rich oxide film reacting with CO at
3
steady state through a red-ox process. However, under CO-rich conditions, a dewetting of
the film occurs that causes the catalyst’s deactivation.
The paper is organised as follows. First, we show results on CO oxidation and
structural characterization of the FeO films at stoichiometric CO:O2 ratios. Then, we provide
data on reactivity at different CO and O2 ratios and partial pressures. Finally, we speculate
on a general mechanism for the CO oxidation reaction on ultrathin FeO films.
2. Experimental
The experiments were performed in three UHV chambers (“TPD-GC”, “STM” and
“PM-IRAS”), equipped with low energy electron diffraction (LEED), Auger electron
spectroscopy (AES) and a quadrupole mass spectrometer (QMS). The TPD-GC chamber
houses a high-pressure cell (~ 30 ml, made of Au-plated Cu massive block) connected to gas
handling lines and a gas chromatograph GC 6890N (Agilent). The double-side polished Pt
(111) crystal (~ 10 mm in diameter, 1.5 mm in thickness) was spot-welded to two parallel Ta
wires, which were in turn welded to two Ta sticks used for resistive heating and also for
cooling by filling a manipulator rod with liquid nitrogen. The temperature was measured by
a chromel-alumel thermocouple spot-welded to the edge of the crystal and controlled using a
feedback system (Schlichting Phys. Instrum.). The manipulator rod inside the chamber ends
with a KF-type flange with a 4-pins electrical feedthrough holding Ta and thermocouple
sticks. The reactor is sealed from the UHV chamber using a Viton O-ring placed on top of
the reactor matching the flange on the rod.
For high-pressure experiments, CO (99.995%, Linde) and O2 (99.999%, AGA
GmbH) were additionally cleaned using a cold trap kept at ~ 200 K. The reaction mixtures
of CO and O2 were balanced by He to 1 bar in a gas handling system. After introductio n to
the high pressure cell, the gas was circulating with a flow of 3 ml/min for 20 min at room
temperature to equilibrate the reaction gas flow. During this pretreatment, no CO2 formation
was observed. Then the sample was heated up to the reaction temperature with a heating rate
of 1 K/s. The gas composition in the circulating flow was analyzed using a HP -Plot Q
column at 35°C and a TCD detector. For structural characterisation of the spent catalysts, the
crystal was rapidly (within 2-3 minutes) cooled down to room temperature, and the reactor
4
was pumped out down to ~10-5 mbar (which typically takes ~ 20 min) before exposing to
UHV.
Basically, a similar design was used in the PM-IRAS chamber, where a single-side
polished Pt(111) crystal was spot -welded by Ta wires to Mo rods on the manipulator for
resistive heating. The temperature was controlled by a chromel-alumel thermocouple spotwelded to the backside of the crystal. After surface preparation and characterization, the
sample was transferred into a stain less steal high-pressure cell (~ 1 l) sealed with
differentially pumped, spring loaded Teflon O-rings. The cell was equipped with two CaF2
windows for infrared reflection absorption spectroscopy (IRAS) studies using a Bruker IFS
66v spectrometer. Polarisation modulated (PM) IRAS measurements were carried out with a
wire-grid polarizer and a photoelastic modulator (Hinds Instruments PEM 90) which
modulates the polarisation of the incident infrared light between p- and s-polarized at a
frequency of 74 kHz. In p-polarisation, both surface and gas-phase species contribute to the
absorption signal, while in s-polarisation, only gas-phase species are detected. By
calculating the differential absorbance ∆R/R = (Rp -Rs)/(Rp +Rs), where indexes s and p are
referred to s- and p-polarisations, respectively, a vibrational spectrum of the surface species
can be obtained.
In the STM chamber, the Pt(111) crystal was mounted to a Pt sample holder. The
temperature was controlled using a chromel-alumel thermocouple spot-welded to the edge of
the crystal. The crystal can be heated in the UHV chamber by electron bombardment from
the backside using a tungsten filament. For treatments at high pressures the sample was
transferred into the Au-plated reactor housing a heating stage, consisting of ceramic and
sapphire pieces. The sample was heated from the backside using a halogen lamp. The STM
images presented here were obtained at tunneling currents of 0.5 – 0.8 nA and positive
sample bias of 0.2 – 1 V.
The preparation of the ultra-thin FeO(111) and nm-thick Fe3 O4 (111) films on
Pt(111) is described elsewhere [18]. Briefly, one monolayer (ML) of Fe (99.95%,
Goodfellow) is deposited onto clean Pt(111) at 300 K and subsequently annealed in 10-6
mbar O2 at 1000 K for 2 min. Repeated cycles of 5 ML Fe deposition and oxidation results
in well ordered Fe3 O4 (111) films.
5
3. Results and discussion
Before discussing reactivity of model catalysts at elevated pressures, it is instructive
here to briefly summarize the results obtained by TPD and TPR for adsorption and coadsorption of CO and O2 under UHV conditions.
The clean Pt(111) surface exposed to saturating amounts of O2 (typically, 20
Langmuirs (L), 1 L = 10-6 Torr s) and subsequently CO at 100 K showed a broad CO2
desorption signal centred at 315 K. In the opposite sequence or exposing the sample to the
stoichiometric mixtures of CO and O2 (2 : 1) at 100 K, no CO 2 was observed in the TPR
spectra. These results are consistent with the well-established CO poisoning effect on
oxygen dissociation.
On the FeO(111)/Pt(111) surfaces, the individual TPD spectra of CO and O2
inversely scale with the FeO coverage, with no difference in the temperature profiles as
compared to the clean Pt(111), thus indicating no specific adsorption at the rim of FeO(111)
islands formed at sub-monolayer coverage. As expected, the FeO(111) films fully covering
Pt(111) do not show CO2 formation in TPR spectra, since neither CO nor O2 adsorbs on the
O-terminated FeO(111) surface at the temperatures used. Note also that long exposures of
the FeO films to 10-6 mbar CO at 500 K did not reveal any signature of oxide reduction. The
results basically confirm the previous conclusions [18] about good thermal stability and
chemical inertness of the FeO(111) films. However, the situation changes dramatically in the
mbar-pressure range.
3.1. Structure and reactivity at stoichiometric CO : O2 ratios
Figure 1 shows CO2 evolution upon slow heating of FeO(111) films in the mixture of
40 mbar CO and 20 mbar O2 balanced by He to atmospheric pressure. Carbon dioxide is
clearly observed at temperatures above 430 K. The reaction rate apparently increases at
increasing temperature as shown in Fig. 2a presenting the kinetics of the CO2 formation at
different temperatures. At all temperatures studied, the oxidation rate decreases in time.
For comparison, the clean Pt(111) surface showed almost no activity under the same
conditions (see Fig. 2b), that is consistent with the above TPR results on CO self-poisoning.
6
On the other hand, the well-ordered, nm-thick Fe3 O4 (111) films grown on Pt(111) also
exhibited much lower activity than the FeO films , although higher than of Pt(111).
Therefore, the enhanced activity observed in these experiments must be intimately
connected to the ultrathin FeO(111) film on Pt(111).
Furthermore, Fig. 2b shows that the reactivity of the FeO films strongly depends on
the pressure while keeping the same CO:O2 ratio. The reaction proceeds much slower when
the pressur e is reduced from 60 to 15 mbar, and then to 6 mbar. The latter fact points either
to different reaction orders for CO and O2 or pressure-dependent surface transformations, or
both. Therefore, we first performed a structural characterization of the model catalysts after
reaction.
Figure 3 shows LEED patterns of the FeO/Pt surface as prepared (a) and after
reaction in 60 mbar of a stoichiometric CO + O2 mixture at 400 (b) and 450 K (c). The sixspots ring around integer spots is characteristic for the Moire superstructure that arises from
the mismatch between FeO(111) and Pt(111) lattices [17,18]. No changes are observed by
LEED after the reaction at 400 K indicating that the surface preserves long-range ordering.
The conclusion holds true also for the samples treated in 6 mbar at 450 K (not shown). On
the contrary, the sample reacted at 450 K in 60 mbar of CO + O2 showed the sharp
diffraction spots of Pt(111) with additional weak spots (Fig. 3c), which are practically
identical to those observed on clean Pt(111) by exposure to CO and assigned to a Pt(111)c(4x2)CO structure [30]. At intermediate temperatures (~ 430 K), a superposition of the
FeO(111)/Pt(111) and CO/Pt(111) structures was observed. Therefore, we conclude that the
FeO(111) films undergo strong reconstruction resulting in a surface exposing a significant
fraction of Pt(111).
The AES study of the spent catalysts revealed only small amounts of carbon beyond
the elements, which belong to the original surfaces, as shown in Fig. 4. (In particular, no
nickel via Ni carbonyls which may contaminate CO in high-pressure containers has been
detected). Carbon (but no oxygen) has been found on the post-reacted Pt(111) surface most
likely due to CO dissociation on the Pt low coordination sites [31]. Therefore, ti seems
plausible that carbon on the FeO/Pt surface appears in course of surface reconstructions
whereby CO dissociates on the open Pt(111) areas. In addition, using the Auger O:Fe signal
ratio in the pristine FeO films as a reference, the iron oxide phase in the post-reacted
7
catalysts exhibits a FeOx (x ~ 1.25) stoichiometry, on average. Subsequent UHV annealing at
temperatures above 800 K essentially restores both the LEED pattern and the O:Fe ratio of
the original film and also removes carbon from the surface (see bottom spectrum in Fig. 4).
In line with LEED data, Auger spectra revealed no significant differences in the
surface composition after catalytic tests at lower temperatures and/or lower pressures. The
threshold temperature observed for the struc tural changes (~ 430 K) basically coincides with
the “ignition” temperature in the reaction profile shown in Fig. 1.
To determine the surface morphology of the model catalysts we employed STM.
The Pt(111) surface after reaction essentially showed the original topography. Only few
small, irregularly shaped particles, presumably of carbon (as judged by AES), and primarily
located at the terrace steps were observed [28]. Fig. 5a shows an STM image of a
FeO(111)/Pt(111) film with atomically flat, wide terraces exhibiting Moire superstructure
with a periodicity ~ 2.6 nm. After 10 minutes exposure to the 40 mbar CO + 20 mbar O2
mixture at 450 K, small particles emerge on the surface (see Fig. 5b), whereas the rest of the
surface shows FeO-like superstructure. The large-scale STM images show that the
reconstruction proceeds homogeneously on the whole film. After 120 min the surface is
covered by nanoparticles as shown in Figs. 5c and 6a. The particles exhibit a narrow size
distribution of 8 ± 1 nm in diameter and 1. 8 ± 0.1 nm in height, and are assigned to iron
oxide particles formed on Pt(111) upon dewetting of a FeO film. Stepwise annealing in
UHV results in the oxide re-wetting (see Fig. 6), in full agreement with the LEED and AES
results.
In order to obtain in situ information on surface restructuring we employed PMIRAS that allowed monitoring surface species simultaneously with the gas-phase
composition. Figure 7a shows a series of spectra in the CO stretching region obtained from
the FeO(111)/Pt(111) surface at 450 K in a reaction mixture of 40 mbar CO and 20 mbar O2
(balanced by He). The spectrum for a Pt(111) crystal under the same conditions for 120 min
is also shown for comparison.
The absorption band at ~ 2095 cm-1 that grows with time is characteristic for atop
(or terminal) CO species on Pt(111). In addition, a broad band centred at ~ 1865 cm-1
develops, which is typical for CO occupying bridge position, i.e., consistent with the
formation of c(4x2)-CO overlayer on Pt(111) observed by LEED. The integral intensity of
8
two bands apparently saturates after ~ 60 min of the reaction as shown in Fig. 7b, finally
exposing approximately 70% of Pt(111). Therefore, the PM IRAS results unambiguously
show that an FeO film dewets a Pt substrate under the reaction conditions, thus exposing a
Pt(111) substrate. Interestingly, the major peak at 2095 cm-1 revealed a low frequency
shoulder (centered at ~ 2075 cm-1 ), which remained after sample cooling and pumping the
reactor out. Therefore, this feature can hardly be assigne d to CO adsorbed on Lewis (Fe 2+ )
sites on iron oxide particles as CO desorbs from iron oxide surfaces below room temperature
[32]. On the other hand, this feature is missing in the spectra for Pt(111) under the same
conditions. Therefore, its presence on the dewetted surface could be linked to the
oxide/metal interface formed upon dewetting.
Figure 7b shows the time evolution of integral amounts of CO adsorbed on the
surface and of CO2 produced in the reaction as obtained by the respective surface and gasphase IR signals. The kinetics of the CO2 production resembles that of the GC experiments
presented in Fig. 2, both showing that the CO oxidation rate (the slope of the CO2
production curve) decreases with time. It is clearly seen that the highest rate is observed at
the beginning of the reaction where the surface is essentially non-dewetted (see more in the
next section). With increasing degree of dewetting the reaction rate slows down and stays
practically constant as the maximum dewetting is obtained. At this stage, the CO2 formation
rate is almost equal to that observed for a Pt(111) single crystal (not shown here). Note that,
under the conditions studied, the CO conversion is very low (< 1%); therefore we can
neglect the changes in the gas composition during the reaction.
Further information on the structure of the post-reacted FeO/Pt surfaces was
obtained by TPD. Only CO and CO2 were found as desorbing species upon heating of the
spent catalysts in UHV. After reaction at low temperatures, a broad and rather featureless
CO desorption signal at 350 - 450 K is observed (Fig. 8). (Note, that the pristine FeO films
do not adsorb CO even at 90 K). With increasing the reaction temperature, a peak at ~ 390 K
develops which superimposes with the signal at 320 - 500 K gaining intensity. Except the
prominent feature at 380-390 K, the TPD spectra are similar to that observed after CO
adsorption on clean Pt(111) shown in the same graph for comparison. The desorption traces
of CO and CO2 at T > 500 K, which are missing in the spectra for Pt(111), may be attributed
to the reaction of carbonaceous species with surface oxygen (e.g., see [39]). This oxygen is
9
apparently coming from the iron oxide particles since the carbon peak in AES spectra
disappears upon heating to 800 K with a simultaneous restoration of the Auger O:Fe ratio
(see Fig. 6). Annealing at T > 800 K leads again to the chemical inertness towards CO, as it
essentially recovers the structure of the FeO(111) films as judged by LEED, AES and STM.
Regarding the desorption peak at ~ 380 K, it cannot be assigned to CO adsorbed on
iron oxide particles [32]. Note that this peak was only observed on strongly dewetted
surfaces (e.g., after reaction in 60 mbar at 450 K) and never under O-rich reaction conditions
(see below). Also, the peak does not show up in the next CO TPD runs, i.e., after the first
heating to 600 K. On the other hand, narrow CO desorption signals are typical for
decomposition of metal carbonyls (e.g., [33, 34]). Indeed, Ptx (CO)y species prepared on
Pt(111) by physical vapour deposition of Pt in 10-6 mbar of CO at 100 K revealed a similar
TPD peak at ~ 380 K superimposed with the signal from CO/Pt(111) as shown in Fig. 8.
Therefore, we may assign this desorption feature, observed on the spent catalysts, to Pt
carbonyl-like species formed upon film dewetting the Pt surface. Its absence on the postreacted Pt(111) sample indicates that these species are probably stabilized by the iron oxide
particles formed. This conclusion is also consistent with the observation of a low frequency
shoulder on the 2095 cm-1 band on FeO/Pt and not on Pt(111) surface in the PM-IRAS
experiments (see Fig. 7a).
One could, in principle, suggest the formation of iron carbonyls upon reduction of
the FeO film at high CO pressures. However, previous studies showed that Fex (CO)y species
(x= 1 – 3; y= 5-12) exhibit the 2064 cm-1 band as the highest for the terminal CO stretch
(e.g., [34-36]), which is well below than the shoulder at 2075 cm-1 observed in Fig. 7a.
Formation of highly reduced or metallic Fe particles adsorbing CO can be excluded either.
Indeed, CO on oxide supported Fe nanoparticles showed the band at ~ 2030 cm-1 [37]. In
addition, if formed upon reduction of the FeO film, the Fe overlayer on Pt(111) would
manifest itself by the IRAS signal either at ~1950 cm-1 (at 343 K) or at 2060 cm-1 (at 473 K),
the latter being due to Fe migration into the sub-surface region as reported in [38]. All
scenarios are inconsistent with the spectra shown in Fig. 7a.
In addition to PM-IRAS, we used the TPD results for estimating the degree of
dewetting (determined here as a percentage of the Pt(111) substrate that opens during a
reaction) by integrating the CO TPD signal between 300 and 500 K. Using the TPD
10
spectrum for CO/Pt(111) as a reference, we found that approximately 70 % of the surface
expose the Pt(111) after reaction at 450 K for 2 h. This value nicely agrees with that
obtained by PM-IRAS. It has turned out that the integral CO intensity vs reaction
temperature relationship fits well the Arrhenius plot as shown in the inset in Fig. 8, from
which the activation energy ~ 70 kJ/mol is calculated. The analysis with a proper subtraction
of the feature at ~ 380 K in the TPD spectra results in a bit smaller value, ~ 65 kJ/mol.
Therefore, 65 - 70 kJ/mol may be considered in the first approximation as the activation
energy for the FeO film dewetting; however, this value will depend on pressure and CO:O2
ratio.
Note, that according to the TPD studies, no dewetting is observed after individual
exposures of FeO films to 40 mbar CO or 20 mbar O2 at 450 K. In both cases, the LEED
patterns revealed a Moire superstructure. It is therefore clear that dewetting is a reactioninduced process.
3.2. O2 -rich vs CO-rich reaction conditions
It is well documented in the literature that the CO oxidation rate on Pt catalysts
strongly depends on CO:O2 partial pressures ([39,41] and references therein). Therefore, one
may raise the question about whether the promoting effect of the ultrathin FeO films exists
in a wide range of CO and O2 partial pressures. To study possible effects of the gas
composition on reactivity of the FeO films, we carried out two sets of experiments in various
CO + O2 mixtures. In the first set, we varied the O2 pressure while keeping the CO partial
pressure at 10 mbar. In the second one, we kept the O2 partial pressure constant at 20 mbar
and varied the CO pressure. In two sets the CO:O2 ratio was varied between 1:5 and 5:1
which basically covers both O2 - and CO-rich regimes. The mixtures were always balanced
by He to 1 bar, and the reaction was monitored at 450 K. After reactivity measurements, the
surfaces were characterized by LEED, TPD and AES.
Figure 9 shows the kinetics of CO2 production at different CO and O2 partial
pressures. Under CO-lean conditions the reaction rate is almost constant in time and
increases while approaching 100 % conversion of CO. In contrast, the reaction slows down
in the CO-rich conditions. Note, that the reaction rate was not much affected by CO2
11
accumulation in the circulating gas mixture. Pumping out the reactor and refilling the line
with a fresh CO + O2 mixture had basically no effect on the subsequent activity, thus
indicating that CO2 is not involved in the deactivation process. For further analysis we
present the rates measured in the first minutes of reaction (i.e., extrapolated to zero
conversion on yet non-deactivated catalysts), if not specified.
Figure 10a depicts the reaction rate as a function of oxygen pressure. The reaction
kinetics exhibits first order (n = 1.1 ± 0.1) for O2 . This may straightforwardly explain the
results of Fig. 2b showing much lower CO2 production upon decreasing the total CO + O2
(2:1) pressure from 60 down to 6 mbar. The reaction order measured on Pt(111) under the
same conditions is close to unity (n = 0.85 ± 0.1) and is consistent with the previous studies
[42]. The figure shows that the FeO/Pt catalysts exhibit much higher activity than Pt(111) in
the whole pressure range studied.
The temperature dependence of the reaction rate measured at 410 – 450 K in the
mixture of 10 mbar CO and 20 mbar O2 leads to an Arrhenius plot with an apparent
activation energy of 113 ± 5 kJ/mol. Due to the very low conversion, it was impossible to
precisely measure the activation energy on Pt(111) for comparison. Nevertheless, this value
is significantly lower than ~ 135 kJ/mol, reported for supported Pt catalysts and Pt single
crystal surfaces for conditions where CO is the primary surface species [40,42], which is in
turn very close to the desorption energy of CO from Pt [30,31].
The reaction is accompanied by remarkable structural changes detected by AES.
The film becomes enriched with oxygen upon increasing O2 pressure, and approaches the
O:Fe ratio close to 2, on average (see Fig. 10b). Interestingly, the corresponding LEED
patterns (not shown here) are in fact similar to those of the pristine FeO(111)/Pt(111) surface.
Another intriguing finding is that the degree of dewetting (see Fig. 10c) basically shows no
direct relationship to activity (Fig. 10a). In fact, in the O2-rich atmosphere the inverse
relation is observed such that the activity is the highest for the least dewetted surface.
In the next step, we review experimental results of the second set of experiments.
Figure 11 shows the reaction rate (a), average stoichiometry of the iron oxide film (b), and
degree of dewetting (c) as a function of CO partial pressure. Note, that under the same
conditions the Pt(111) surface showed zero-order kinetics for CO at 450 K, i.e., in full
agreement with the previous studies [42]. Figure 11a reveals again, that the FeO films are
12
much more active than Pt(111) in the whole CO pressure range studied. The initial reaction
rate vs CO pressure relationship is not monotonous and has a maximum at CO:O2 ~ 1.
Apparently, at high CO pressures the reaction exhibits negative reaction order, and the
catalytic activity decreases with time (see Fig. 9b and also Fig. 2). The data measured after 2
h on stream are also shown in Fig. 10a, for comparison.
The dewetting process remarkably accelerates at CO pressures above 10 mbar (see
Fig. 11c) leading to a degree of dewetting of ~ 70 % as the upper limit. However, LEED
and TPD inspections of selected samples after 20 - 30 min on stream, i.e., before the
deactivation sets in, showed negligible dewetting. Note also, that in the first set of
experiments the highest activity is observed for the non-dewetted surfaces. Finally, the PMIRAS results (see Fig. 7b) showed in situ that the reaction rate is the highest in the first
minutes of reaction. Therefore, we conclude that dewetting in fact causes the catalyst’s
deactiviation rather than forming the most active phase as originally proposed [28].
Finally, under O-rich conditions, the iron oxide becomes enriched with oxygen.
Interestingly, O-enrichment apparently mirrors the degree of dewetting (see Figs. 11(b,c)).
3.3. General discussion and proposed mechanism
The key observations presented in the previous sections can be summarized as
follows.
(i)
FeO(111) monolayer films grown on Pt(111) show much higher CO
oxidation activity than Pt(111) at 400 - 450 K in a wide range of CO:O2
ratios (between 1:5 and 5:1). The effect is observed only in the mbar
pressure range.
(ii)
The reaction is accompanied by strong structural transformations. In O2 rich ambient, the films become enriched with oxygen while maintaining
the long range ordering. Under CO-rich conditions, the films undergo
dewetting that causes the catalyst’s deactivation. Dewetting is a reaction
induced process and cannot be observed in pure O2 or CO ambient.
(iii)
The reaction rate at 450 K exhibits first order for O2 and nonmonotonously depends on CO pressure.
13
In addition, it appears that the reaction does not exhibit an induction period, or the
latter is shorter than our acquisition time (~ 5 min) (see Fig. 2 and Fig. 9). Under O-rich
conditions, the film is enriched with oxygen and does not undergo dewetting that in turn
slows down the reaction rate in the CO-rich atmosphere (see Figs. 9-11). These findings
suggest that the surface of the O-rich film formed at elevated O2 partial pressures is in fact
the most active. The surface transformations resulting in this surface seem to be relatively
fast since the reaction rate in the O-rich atmosphere is practically constant from the onset.
The lack of structural information of the reconstructed FeOx film surface renders the
reaction mechanism uncertain. Oxidation reactions on oxides are usually considered within a
Mars-van Krevelen scheme [43] which invokes a surface redox process where the reactant is
oxidized by lattice oxygen and the reduced catalyst is subsequently reoxidized by molecular
oxygen. However, the recent analysis by Vannice [44] revealed a number of inconsistencies
in the assumptions incorporated into the derivation of the rate expression, even though still
the experimental data may be fitted.
Oxidation of CO on iron oxides (Fe 2 O3) has been previously studied both
experimentally and theoretically [45-48]. The reaction is typically carried out at
temperatures above 500 K in excess of oxygen, although the reaction may occur even in the
absence of external oxygen over iron oxide nanoparticles [45]. It was found that the reaction
proceeds through dissociative adsorption of O2, which is considered to be a non-activated
process, and removal of oxygen by a gas-phase CO via an Eley-Rideal type mechanism.
The reaction showed zero-order kinetics for O2 and first order for CO at 520 – 570 K
[44,45]. Microkinetic modeling revealed that the reaction between CO and surface O species
is the rate-determining step [46]. The Eley-Rideal mechanism was corroborated theoretically
for the (100) surface of Fe 2 O3, whereas on the oxygen terminated (0001) surface the
preference towards the Langmuir-Hinshelwood mechanism (i.e., CO adsorption before
reaction with O) was observed [47].
Certainly, in these studies the bulk structure of iron oxide was assumed to remain
unchanged during reaction, whereas the monolayer FeO films undergo strong
transformations, which should be taken into account in the reaction kinetics. Our results
showed first order kinetics for O2 , i.e., in variance to the above-mentioned iron oxides
14
catalysts showing zero order. Therefore, it may well be that the oxygen pressure dependence
reflects the kinetics of structural transformations resulting in the O-rich phase rather than the
reaction itself. It seems plausible that the whole reaction starts with the formation of the Orich surface domains, which react with CO producing CO2. The lattice oxygen is then
replenished from the gas phase. In excess of CO the formation of the O-rich film competes
with the film reduction accompanied by dewetting.
The atomic structure of the O-rich surface is still unclear. However, we may
speculate on possible structures. The FeO films, either treated by pure O2 or reacted at low
CO:O2 ratios at 450 K, showed LEED patterns, which are very similar to those of the
pristine FeO(111)/Pt(111) surface. On the other hand, these films accommodate much more
oxygen (without loosing Fe), thus approaching the average Fe:O ratio clos e to 1:2 as judged
by AES (see Fig. 10b). Therefore, we have tentatively proposed that a FeO film, originally
composed of Fe and O close-packed layers stacked as O-Fe/Pt, reconstructs at high O2
pressures into a O-Fe-O/Pt –like (“sandwich”) structure. Inde ed, STM images of the
FeO(111) films treated with 20 mbar of O2 at 450 K showed long-range modulation which is
very similar to that of pristine FeO (Fig. 12).
Certainly, whenever reactions at elevated pressures are carried out, very careful
attention needs to be considered towards impurities in the feedstock, particularly to those
having high reaction probability. There are evidences in the literature showing that oxide
surfaces may be very sensitive to the traces of water in an ambient, even in the vacuum
background [49,50].
In order to see whether water and hydrogen as impurities in the reaction ambient
affect the reactivity data presented in Figs. 9-11, we have performed experiments in the
TPD-GC setup using the mixture of 10 mbar CO + 50 mbar O2 with additional 0.5 mbar of
water. The results showed no effect of water on reactivity under these conditions.
Meanwhile, adding 0.5 mbar of H2 to a CO + O2 mixture dramatically increased the reaction
rate, by factor of 2. However, in the latter case the FeO(111) film reconstructs in a different
manner, and the reaction most likely proceeds through the different mechanism (to be
discussed in forthcoming paper).
In addition, we have examined thermal stability of the O-rich FeO x films formed by
exposure to 20 mbar O2 at 450 K for 20 min. The TPD spectra on heating to 1200 K (not
15
shown) showed ~ 90 % increase of oxygen content as compared to the pristine film, i.e.
consistent with the AES results presented in Figs. 10-11, but did not reveal water desorption.
Only tiny amounts of water, as calibrated by water TPD on pristine FeO [51], was found
desorbing in a broad signal at 400 - 700 K upon adding 0.15 mbar of H2 into 20 mbar of O2 .
Therefore, the TPD results suggest no hydroxyls to be present on the O-rich film surface
under reaction conditions.
On the other hand, IRAS study of the FeO films exposed to 20 mbar of O2 at 300 K
in the PM-IRAS setup, having much higher reactor volume (~ 1 l vs 30 ml in GC setup),
revealed an absorption band at 3642 cm-1 that shifts to 3625 cm-1 upon heating to 450 K.
This band corresponds to the stretching mode of “isolated” (not H-bonded) OH species. The
band stayed basically unchanged after pumping oxygen out and disappeared upon heating in
vacuum above 500 K, accompanied by water desorption clearly seen in TPD spectra. The
unexpected observation of surface hydroxyls on the O2 -treated films can be explained either
by impurities in the feedstock or those are formed by reactions on the stainless steal reactor
walls and gas handling lines. These species were not observed in 10 mbar of pure CO and 1
mbar of pure H2 , and only with minor abundance in 1 mbar of water. Note, that the latter
amounts are far above those expected as impurity in the reactor filled with 20 mbar of O2 .
These findings therefore suggest that OH species are in fact formed as a result of the
reaction of H2 or water with the film first formed upon interaction of FeO with O2 . It appears
that O-rich film is extremely reactive towards traces of hydrogen (or water) in the
background, ultimately forming OH species. The resulted film is probably reminiscent of
iron oxyhydroxides. Whether this may be connected to results on reduction of FeO films
with atomic hydrogen [52, 53] is not clear yet.
4. Summary
In this paper, we have shown that a thin FeO film exhibits enhanced activity towards
CO oxidation as compared to clean Pt(111) and the nm-thick Fe 3O4 (111) films in a wide
range of CO:O2 ratios at 450 K. Post -characterization of the model catalysts using STM,
LEED, AES and TPD shows , that the reaction most likely proceeds through the formation of
an oxygen-rich FeOx (1 < x < 2) film that reacts with CO via the redox mechanism. The O-
16
rich film tends to dewet at CO:O 2 > 1, ultimately resulting in highly dispersed iron oxide
particles on Pt(111). The film dewetting was monitored in situ by PM-IRAS. The
wetting/dewetting process is reversible depending on the ambient conditions.
Such strong restructuring of the O-terminated FeO surface, in particular the formation
of the well-ordered, oxygen-rich film at elevated O2 pressures, implies an enhanced reactivity
of ultrathin films towards gas phase molecules, otherwise inert at low pressures and/or on
thick oxide films. This situation apparently links the reactivity of a thin oxide film to the
presence of a metal substrate underneath and electron transfer processes recently highlighted
in the literature [21-27]. In turn, the results may aid in our deeper understanding of the
reactivity of metal particles encapsulated by oxide films as a result of strong metal support
interaction, e.g., Pt/TiO 2 , Pt/CeO2 , Pt/Fe3 O4 , etc. Vibrational and electron spectroscopy
studies with the goal to identify the nature of the oxygen rich phase are underway.
Acknowledgments.
We acknowledge support from the Deutsche Forschungsgemeinschaft (DFG)
through SFB546 (“Structure and reactivity of transition metal oxides”) and the Cluster of
Excellence “Unifying concepts in catalysis” (UNICAT), coordinated by TU Berlin, and the
Fonds der Chemischen Industrie. Y.-N.S thanks International Max Planck Research School
“Complex Surfaces in Materials Science”. E.C. thanks the Spanish Ministry of Science and
Innovation for a fellowship.
17
References
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
[13]
[14]
[15]
[16]
[17]
[18]
[19]
[20]
[21]
[22]
[23]
[24]
[25]
[26]
[27]
[28]
[29]
[30]
[31]
[32]
S.J. Tauster, S.C. Fung and R.L. Garten, J. Am. Chem. Soc. 100 (1978) 170.
E. Ko, R. Garten, J. Catal. 68 (1981) 223.
S.J. Tauster, Acc. Chem. Ress. 20 (1987) 389.
F. Solymosi, J. Catal., 94 (1985), 581.
G.L. Haller, D.E. Resasco, Adv. Catal. 36 (1989) 173.
F. Pesty, H.P. Steinrück, T.E. Madey, Surf. Sci. 339 (1995) 83.
O. Dulub, W. Hebenstreit, U. Diebold, Phys. Rev. Lett. 84 (2000) 3646.
M. Bowker, P. Stone, P. Morrall, R. Smith, R. Bennett, N. Perkins, R. Kvon, C. Pang,
E. Fourre, M. Hall, J. Catal. 234 ( 2005) 172.
F. Silly, M.R. Castell, J. Phys. Chem. B 109 (2005) 12316.
K.J. Blankenburg, A.K. Datye, J. Catal. 128 (1991) 186.
A.K. Datye, D.S. Kalakkad, M.H. Yao, D.J. Smith, J. Catal. 155 (1995) 148.
S. Bernal, J.J. Calvino, M.A. Cauqui, J.M. Gatica,J.A. Perez Omil, J.M. Pintado,
Catal. Today 50 (1999) 175.
D.R. Mullins, K.Z.Zhang, Surf. Sci., 513(2002), 163.
S. Surnev, J. Schoiswohl, G. Kresse, M.G. Ramsey, F.P. Netzer, Phys. Rev. Lett. 89
(2002) 246101.
Z.H. Qin, M. Lewandowski, Y.N. Sun, S. Shaikhutdinov, H.-J. Freund, J. Phys. Chem.
C. 112 (2008) 10209.
Z.H. Qin, M. Lewandowski, Y.N. Sun, S. Shaikhutdinov, H.-J. Freund, J. Phys.:
Condens. Matter, 21 (2009) 134019.
H.C. Galloway, P. Sautet, M. Salmeron, Phys. Rev. B 54(1996), R11145.
W. Weiss, W. Ranke, Progr. Surf. Sci., 70 (2002) 1.
R. S. Cutting, C.A. Muryn, G. Thornton, D.J. Vaughan, Geochim. Cosmochim. Acta,
79 (2006) 3593.
K. Abib, D.R. Mullins, G. Totir, N. Camillione, J.P. Fitts, K.T. Rim, G.W. Flynn,
R.M. Osgood, Surf. Sci. 537 (2003) 191.
G. Pacchioni, L. Giordano, M. Baistrocchi, Phys. Rev. Lett. 94 (2005), 226104.
H.-J. Freund, G. Pacchioni, Chem. Soc. Rev. 37 (2008), 2224.
H.-J. Freund, Surf. Sci. 601 (2007), 1438.
M. Sterrer, T. Risse, U. M. Pozzoni, L. Giordano, M. Heyde, H. P. Rust, G.
Pacchioni, H.-J. Freund, Phys. Rev. Lett. 98 (2007), 096107.
M. Sterrer, T. Risse, M. Heyde, H.-P. Rust, H.-J. Freund, Phys. Rev. Lett. 98 (2007),
206103.
H. Grönbeck, J. Phys. Chem. B 110 (2006), 11977.
P. Frondelius, A. Hellman, K. Honkala, H. Hakkinen, H. Grönbeck, Phys. Rev. B 78
(2008), 085426.
Y.-N. Sun, Z.H. Qin, M. Lewandowski, S. Shaikhutdinov, H.J. Freund, Catal. Letters,
126 (2008), 31.
K. Hayek, M. Fuchs, B. Klötzer, W. Reichl, G. Rupprec hter, Topics Catal. 13 (2000)
55.
G. Ertl, M.Neumann, K.M. Streit, Surf. Sci. 64 (1977) 193.
K. McCrea, J. Parker, P. Chen, G. Somorjai, Surf. Sci., 494 (2001) 238.
C. Lemire, R.Meyer, V.E. Nenrich, S. Shaikhutdinov, H.-J. Freund, Surf. Sci., 572
(2004) 103.
18
[33]
[34]
[35]
[36]
[37]
[38]
[39]
[40]
[41]
[42]
[43]
[44]
[45]
[46]
[47]
[48]
[49]
[50]
[51]
[52]
[53]
J. Libuda, A.Sandel, M. Bäumer, H. -J. Freund, Chem. Phys. Lett., 240 (1995) 429.
S. Sato, Y. Ukisu, Surf. Sci. 283 (1993) 137.
S. Fletcher, M. Poliakoff, J.J.Turner, Inorg. Chem. 25 (1986) 3597.
R.L. Jackson, M.R. Trusheim, J. Am. Chem. Soc. 104 (1982) 6590.
M P. Felicissimo, O.N. Martyanov, T. Risse, H.-J. Freund, Surf. Sci. 601 (2007)
2105
T. Wadayama, H. Osano, T. Maeyama, H. Yoshida, K. Murakami, N. Todoroki, and
S. Oda, J. Phys. Chem. C 112 (2008) 8944.
L. Schäfer, H.-W.Wassmuth, Surf. Sci., 208 (1989) 55.
A.K. Santra, D.W. Goodman, Electrochim. Acta 47 (2002) 3595.
R. Imbihl, G. Ertl, Chem. Rev. 95 (1995) 697.
P.J. Berlowitz, C.H.F. Peden, D.W. Goodman, J. Phys. Chem. 90 (1988) 5213.
P. Mars, D.W. van Krevelen, Chem. Eng. Sci., 3 81954) 41.
M.A. Vannice, Catal. Today 123 (2007) 18.
P. Li, D.E. Miser, S. Rabiei, R.T. Yadav, M.R. Hajaligol, Appl. Catal. B 43 (2003)
151.
S. Wagloehner, D. Reichert, D. Leon-Sorzano, P. Balle, B. Geiger, S. Kureti, J. Catal.,
260 (2008) 305.
A.K. Kandalam, B. Chatterjee, S.N.Khanna, B.K. Rao, P. Jena, B.V. Reddy, Surf.
Sci., 601 (2007) 4873.
H. Randall, R. Doepper, A. Renken, Ind. Eng. Chem. Res., 36 (1997) 2996.
O. Bikondoa, C.L. Pang, R. Ithnin, C.A. Muryn, H. Onishi, G. Thornton, Nature
Materials, 5 (2006), 189-192.
R. Schaub, E. Wahlström, A. Rønnau, F. Besenbacher, Science 299 (2003), 377;
Science 314 (2006), 925.
J.L. Daschbach, Z. Dohnalek, S.-R. Liu, R.S. Smith, B.D. Kay, J. Phys. Chem. B 109
(2005), 10362.
W. Huang, W. Ranke, Surf. Sci., 600 (2006), 793.
L.R. Merte, J. Knudsen, L.C. Grabow, R.T. Vang, E. Lægsgaard, M. Mavrikakis, F.
Besenbacher, Surf. Sci. 603 (2009), L15.
19
Figure captions.
Fig. 1. Temperature programmed reaction profile of CO2 production over FeO(111)/Pt(111)
on slow heating (1 K/min) in 40 mbar CO + 20 mbar O2 balanced by He.
Fig. 2. Kinetics of CO2 production over FeO(111)/Pt(111) in: (a) 40 mbar CO + 20 mbar O2
at indicated temperatures; (b) at stoichiometric CO:O2 (2:1) ratios at indicated total pressures
at 450 K. The results for Pt(111) and nm-thick Fe 3 O4(111)/Pt(111) film in 40 mbar CO + 20
mbar O2 at 450 K are shown in (b), for comparison. Time zero corresponds to the start of the
sample heating (1 K/s) from 300 K.
Fig. 3. LEED patterns of the “as prepared” FeO(111)/Pt(111) film (a) and after reaction at
400 (b) and 450 K (c) in 40 mbar CO + 20 mbar O2 in He. The unit cell for c(4x2)CO Pt(111) structure is indicated.
Fig. 4. AES spectra of the clean FeO(111)/Pt(111) film (top), after the reaction in 40 mbar
CO + 20 mbar O2 at 450 K (middle) and subsequent annealing to 800 K for 2 min (bottom).
The corresponding peak ratios for O(510 eV) and Fe(658 eV) are indicated.
Fig. 5. STM images (size 200 nm x 200 nm) of the as prepared FeO(111) film (a), after 10
minutes (b) and two hours (c) in 40 mbar CO + 20 mbar O2 at 450 K. The inset shows the
atomic structure of the Moire pattern.
Fig. 6. STM images (size 500 nm x 500 nm) of the dewetted FeO(111) film (a) and after
step-wise UHV annealing at 500 (b), 700 (c) and 800 K (d).
Fig. 7. (a) Time-resolved PM-IRAS spectra of the CO stretching region obtained from the
FeO(111)/Pt(111) surface at 450 K in a mixture of 40 mbar CO and 20 mbar O2 balanced by
He to 1 bar. The spectrum for the Pt(111) surface after 120 min at the same conditions is
shown for comparison. (b) Kinetics of CO2 formation in the gas phase and of CO adsorbed
on FeO(111)/Pt(111) obtained by integration of the respective PM-IRAS signals.
20
Fig. 8. CO (28 amu) and CO 2 (44 amu) signals in TPD spectra of a FeO(111)/Pt(111) film
reacted in 40 mbar CO + 20 mbar O2 for 120 min at indicated temperatures. The samples
were cooled to 300 K and evacuated to 10-5 mbar before exposing to UHV. The inset shows
the Arrhenius plot for the integral CO desorption signal as a function of reaction temperature.
The spectrum for clean Pt(111) after reaction at 450 K is shown as dotted line for
comparison. TPD spectrum of Ptx (CO) y formed on Pt(111) by Pt deposition in 10-6 mbar of
CO at 100 K (dashed line) is shown to highlight a desorption feature at ~ 380 K.
Fig. 9. Kinetics of CO2 production at 450 K at different CO and O2 partial pressures as
indicated. Full conversion can be reached at low CO:O2 ratios. Note the reaction rate slows
down in CO-rich conditions (CO:O2 >1).
Fig. 10. Reaction rate (a), average stoichiometry (b), degree of dewetting (c) of the spent
FeO/Pt catalysts as a function of O2 pressure in the mixture with 10 mbar of CO, balanced
by He. Reaction temperature is 450 K.
Fig. 11. Reaction rate (a), average stoichiometry (b), degree of dewetting (c) of the spent
FeO/Pt catalysts as a function of CO pressur e in the mixture with 20 mbar of O2, balanced
by He. The reaction rates measured after 120 min on stream as well as data for clean Pt(111)
are also shown.
Fig. 12. STM image (50 nm x 50 nm) of the FeO film treated in 20 mbar of O2 at 450 K for
10 min, showing that the film basically maintains long-range ordering (compare Fig. 5a).
21
Fig. 1
22
Fig. 2
23
Fig. 3
Fig. 4
24
Fig. 5
Fig. 6
25
Fig. 7
Fig. 8
26
Fig. 9
27
Fig. 10
28
Fig. 11
29
Fig. 12
30